Liver histology and diffusion-weighted MRI in children with nonalcoholic fatty liver disease: A MAGNET study

Abstract

Purpose: To determine potential associations between histologic features of pediatric nonalcoholic fatty liver disease (NAFLD) and estimated quantitative magnetic resonance diffusion-weighted imaging (DWI) parameters.

Materials and methods: This prospective, cross-sectional study was performed as part of the Magnetic Resonance Assessment Guiding NAFLD Evaluation and Treatment (MAGNET) ancillary study to the Nonalcoholic Steatohepatitis Clinical Research Network (NASH CRN). Sixty-four children underwent a 3T DWI scan (b-values: 0, 100, and 500 s/mm² ) within 180 days of a clinical liver biopsy of the right hepatic lobe. Three parameters were estimated in the right hepatic lobe: apparent diffusion coefficient (ADC), diffusivity (D), and perfusion fraction (F); the first assuming exponential decay and the latter two assuming biexponential intravoxel incoherent motion. Grading and staging of liver histology were done using the NASH CRN scoring system. Associations between histologic scores and DWI-estimated parameters were tested using multivariate linear regression.

Results: Estimated means ± standard deviations were: ADC: 1.3 (0.94-1.8) × 10^-3 mm² /s; D: 0.82 (0.56-1.0) × 10^-3 mm² /s; and F: 17 (6.0-28)%. Multivariate analyses showed ADC and D decreased with steatosis and F decreased with fibrosis (P < 0.05). Associations between DWI-estimated parameters and other histologic features were not significant: ADC: fibrosis (P = 0.12), lobular inflammation (P = 0.20), portal inflammation (P = 0.27), hepatocellular inflammation (P = 0.29), NASH (P = 0.30); D: fibrosis (P = 0.34), lobular inflammation (P = 0.84), portal inflammation (P = 0.76), hepatocellular inflammation (P = 0.38), NASH (P = 0.81); F: steatosis (P = 0.57), lobular inflammation (P = 0.22), portal inflammation (P = 0.42), hepatocellular inflammation (P = 0.59), NASH (P = 0.07).

Conclusion: In children with NAFLD, steatosis and fibrosis have independent effects on DWI-estimated parameters ADC, D, and F. Further research is needed to determine the underlying mechanisms and clinical implications of these effects.

Level of evidence: 1 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2017;46:1149-1158.

Keywords: ADC; DWI; NAFLD; QIBs; children; intravoxel incoherent motion; nonalcoholic fatty liver disease; pediatric; quantitative imaging biomarkers.

Figure 1

Univariate associations between DWI parameters (ADC top, D middle, F bottom) and each histologic feature (STE: steatosis, FIB: fibrosis, L-INF: lobular inflammation, P-INF: portal inflammation, BAL: ballooning, and presence/absence of NASH) are shown. The boxplot indicates the 1^st, 2^nd, 3^rd, and 4^th quartiles, and the minimum and maximum value of each DWI parameter. Plots for ordinal variables are annotated with the p-value from the Spearman correlation test. Plots for the categorical variable NASH (presence/absence of NASH) are annotated with the p-value from the Wilcoxon signed-rank test.

Figure 2

Parametric maps of ADC, D, and F are shown for a subject with steatosis = 0, fibrosis = 0 (top) and a subject with steatosis = 3, fibrosis = 0 (bottom). Note reduced signal intensity in the liver parenchyma for maps of ADC and D in the subject with elevated steatosis.

Figure 3

Parametric maps of ADC, D, and F are shown for a subject with fibrosis = 0 steatosis = 1 (top) and a subject with fibrosis = 4, steatosis = 1 (bottom). Note reduced signal intensity in the liver parenchyma for the F map in the subject with elevated fibrosis. In this example, signal intensity is also reduced in the ADC map, likely because ADC is a composite parameter which encompasses both diffusion and perfusion effects. In this extreme example of fibrosis, the change in perfusion likely affects ADC as well.

Figure 4

Univariable associations between IVIM DWI parameters (D top, F bottom) and each histologic feature (STE: steatosis, FIB: fibrosis, L-INF: lobular inflammation, P-INF: portal inflammation, BAL: ballooning, and presence/absence of NASH) are shown. The boxplot indicates the 1^st, 2^nd, 3^rd, and 4^th quartiles, and the minimum and maximum value of each DWI parameter. Plots for ordinal variables are annotated with the p-value from the Spearman correlation test. Plots for the categorical variable NASH (presence/absence of NASH) are annotated with the p-value from the Wilcoxon signed-rank test.

Figure 5

Bland-Altman analyses of interobserver reprodcibility. Difference in reader 1 and reader 2 (reader 1 – reader 2) is plotted as a function of mean reader 1 and reader 2 analyses for DWI-derived parameters a) ADC, b) D, c) and F. The blue line in the middle of each graph indicates the mean difference. The two blue lines above and below indicate the 95% limits of agreement.